Modeling complex legal cases as a Bayesian network using idioms and sensitivity analysis with the Collins case as a complete example
MetadataShow full item record
The 1968 People v. Collins trial is famous for misuse of probability. Separate pieces of evidence were assumed independent, hence overestimating their combined evidential value. Researchers have explored the use of BNs to perform joint evaluation of evidence in cases such as Collins. However, two systemic problems limit the use of BN modelling of an entire case. The first is lack of a repeatable method for modelling certain types of evidence consistently; the second is the many conditional probabilities that inevitably need to be subjectively estimated, raising questions of validity of the resulting posterior probabilities. We address both problems using the Collins case as an example. The first is addressed by building on the idioms based approach, while the second is addressed using a thorough sensitivity analysis. By exploiting simulation in R, a sensitivity analysis is performed that shifts the attention from the posterior probabilities to the structure of the model. Modeling complex legal cases as a Bayesian network using idioms and sensitivity analysis with the Collins case as a complete example (PDF Download Available). Available from: https://www.researchgate.net/publication/322721949_Modeling_complex_legal_cases_as_a_Bayesian_network_using_idioms_and_sensitivity_analysis_with_the_Collins_case_as_a_complete_example?channel=doi&linkId=5a6b3942458515b2d0539dbe&showFulltext=true [accessed Feb 20 2018].
AuthorsFENTON, NE; DE ZOETE, J; Lagnado, D; 10th International Conference on Forensic Inference and Statistics
- College Publications